Inference of transfer functions and prediction of vessel responses using machine learning

University essay from KTH/Marina system

Author: Yaolin Ge; [2020]

Keywords: ;

Abstract: Our ocean covers more than 70% of our planet’s surface. It is a huge reservoir continuously supplying us with enormousvaluable resources. The autonomous aquaculture, deep sea mining, subsea oil / gas exploitation and marine biology etc.,are the increasingly important driving forces for the development of underwater robotics. For all underwater robotics,navigation and positioning system often endures a challenging problem due to the high attenuation of radio-frequencysignals and the lack of Global Positioning System (GPS). Normally, acoustic navigation is the only way for theunderwater robotics to have the accurate navigation. One core element for the acoustic localisation is the rangeestimation. To provide an accurate range estimate from the underwater vehicles to the fixed reference point, an dropbeaconneeds to be established. In the underwater environment, the acoustic signals suffer from multi-path effects,ambient noises etc., which together with its simplified hardware can impose restrictions on the development of a perfectsystem. This project employs a very simplified hardware to deal with the range estimation problem from the beacon tothe transmitter subjected to the above mentioned issues. The algorithm consists of three main components includingdownsampling, matched filtering and CFAR detection. The downsampling is comprised of three steps such asbasebanding, lowpass filtering and resampling with lower sampling rate. By simulating four different signal typesincluding sinusoidal signal, frequency-modulated signals and M-sequence signal, the sinusoidal waveform is selected tosuit the system’s objective both for simplicity and robustness. A series of tests including tube test, water tank test, nearfieldopen water tests as well as LoLo integration and far-field tests verified and validated the system’s capability toestimate the accurate range (error ≤ 1m) in near field cases (≤ 20 meters). For far-field tests, it proved some furtherimprovements need to be accomplished before it is able to carry out long range missions.

  AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)